A method and system for monitoring a natural resource region

Abstract

The application provides a natural resource area monitoring method and system, relates to the technical field of natural resource monitoring, and generates a spectral feature layer through multi-source remote sensing data covering a monitoring area, synchronously obtains ground surface monitoring device ground surface polarization information and generates a polarization feature map to distinguish natural object and artificial object reflection characteristics; the spectral features of the natural object state change in the spectral feature layer, the key areas of the artificial object reflection characteristics in the polarization feature map are fused, and potential illegal graph patches are located; according to the continuously obtained spectral feature layer and polarization feature map, the spectral feature conversion and artificial object key area expansion are combined to determine whether the potential illegal graph patch continuously changes. After the continuous change is determined, a target illegal graph patch and geographic space information are generated; the associated digital orthographic image and graph patch boundary are fitted to a monitoring video screen, and accurate positioning of the natural resource area illegal graph patch and dynamic identification of the target activity can be realized.

Description

Technical Field

[0001] This application relates to the field of natural resource monitoring technology, and in particular to a monitoring method and system for natural resource areas. Background Technology

[0002] Regional monitoring of natural resources is a core technological means for land management and ecological protection, enabling the timely detection of abnormal resource utilization. This technology is widely used in scenarios such as farmland protection and ecological red line management, and is of great significance for maintaining the sustainable use of natural resources.

[0003] Currently, existing regional monitoring of natural resources mostly employs single optical remote sensing technology. By acquiring multi-temporal remote sensing images and comparing and analyzing spectral changes in land cover, suspected illegal land cover patches can be identified. This method, with its advantages of wide coverage and convenient data acquisition, has become one of the mainstream monitoring methods.

[0004] However, single optical remote sensing technology struggles to effectively distinguish between changes in the state of natural objects and changes caused by human intervention (such as unauthorized excavator operations or the construction of illegal buildings). It is prone to misjudging natural changes as violations or overlooking violations caused by covert human intervention, leading to a high rate of false positives and false negatives in monitoring results. Therefore, existing technologies suffer from insufficient accuracy in monitoring violations in natural resource areas. Summary of the Invention

[0005] The purpose of this application is to provide a monitoring method and system for natural resource areas to address the problem of insufficient accuracy in monitoring violations in natural resource areas in existing technologies.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for monitoring natural resource areas, comprising:

[0007] Based on the acquired multi-source remote sensing data covering the monitoring area, a spectral feature layer is generated. The spectral feature layer is used to characterize the spectral feature differences of natural objects in the monitoring area as they transition from a first state to a second state.

[0008] Simultaneously acquire surface polarization information collected by ground monitoring devices deployed within the monitoring area, and process it to generate a polarization feature map, which is used to distinguish the reflection characteristics of natural objects from those of man-made objects;

[0009] By fusing the spectral features of natural objects undergoing state changes identified from the spectral feature layer with key regions exhibiting reflective properties of man-made objects identified from the polarization feature map, potential violation patches are located within the monitoring area;

[0010] Based on the continuously acquired spectral feature layer and polarization feature map, it is determined whether the potential violation patch has undergone a continuous transition from the first state to the second state, wherein the determination of the transition needs to be based on both the change in spectral features and the expansion of the key area with the man-made object.

[0011] When the continuous change is determined to occur, a target violation patch and its corresponding geospatial information are generated.

[0012] Based on the geospatial information of the target illegal patch, the associated digital orthophoto image and the patch boundary corresponding to the target illegal patch are overlaid into the monitoring video image at the corresponding location, and the target activity is dynamically identified within the patch boundary.

[0013] Optionally, after locating a potential violation patch within the monitoring area, the step of determining whether the potential violation patch undergoes a continuous transition from a first state to a second state further includes:

[0014] Acquire nighttime light data covering the areas of the potential violation patches;

[0015] Analyze the nighttime light data to identify whether there are any artificial light anomaly areas within the boundaries of the potential violation patches that appear from nothing and whose brightness values ​​and distribution ranges remain stable or increase over time.

[0016] The presence of the aforementioned abnormal artificial lighting area is used as an additional verification condition for determining that the continuous change has occurred.

[0017] When a potential violation patch simultaneously meets three conditions—a continuous shift in spectral characteristics, an expansion of a key area containing artificial objects, and the presence of an abnormal area of ​​artificial lighting—a continuous shift from the first state to the second state is determined to have occurred.

[0018] Optionally, the fusion of spectral features of natural objects undergoing state changes identified from the spectral feature layer and key regions with reflective properties of man-made objects identified from the polarization feature map, to locate potential violation patches within the monitoring area, includes:

[0019] By comparing spectral feature layers from different time periods using a change vector analysis algorithm, and calculating the vegetation index that represents the amount of vegetation cover change, pixels with significantly reduced vegetation indexes are identified as preliminary candidate regions for state change.

[0020] Simultaneously, the linear polarization degree is calculated based on the surface polarization information, and pixels with high linear polarization degree are selected by setting a threshold to generate key areas representing the reflective properties of man-made objects;

[0021] By using a weighted decision fusion model, the preliminary candidate region and the key region are spatially overlaid and analyzed. Overlapping regions that simultaneously meet the two conditions of significantly reduced vegetation index and high linear polarization are clustered to locate potential violation patches within the monitoring area.

[0022] Optionally, determining whether the potential violation patch undergoes a continuous transition from a first state to a second state based on the continuously acquired spectral feature layer and polarization feature map includes:

[0023] For each potential violation patch, a spectral and polarization characteristic profile is established over a continuous time series; by performing trend analysis on the mean vegetation index within the patch, it is determined whether the vegetation cover of the potential violation patch shows a downward trend.

[0024] Calculate the area of ​​key regions with reflective properties of man-made objects in the potential violation patches, and determine whether the area is increasing over time.

[0025] A transformation process is considered a continuous transformation if and only if a potential violation patch simultaneously meets both criteria: a continuous decline in vegetation cover and a significant expansion in the area of ​​key man-made structures.

[0026] Optionally, the step of dynamically identifying target activity within the boundary of the patch includes:

[0027] Extract the pixel region located within the boundary of the patch from the continuous surveillance video footage;

[0028] By comparing the pixel value differences in the pixel region in adjacent video frames, the difference region representing the pixel value change is obtained;

[0029] Based on the correlation between the position and shape of the difference region in the image in multiple consecutive frames, at least one independently moving target is determined, and the continuous position of the moving target in the image is tracked.

[0030] Extract the shape, texture and motion pattern features of the moving target in the video frame, and classify the moving target into a predefined dynamic target type based on preset feature comparison rules;

[0031] Analyze the change pattern of the continuous position of the tracked moving target over time, and determine the behavior category of the moving target according to the preset behavior pattern rules;

[0032] When the dynamic target type and at least one of the behavior categories do not meet the preset violation judgment conditions, a task awaiting manual verification is generated.

[0033] Optionally, the generation of the spectral feature layer includes:

[0034] Reflectance information in the red and near-infrared bands, which is sensitive to vegetation status, is extracted from the first and second spectral images of the multi-source remote sensing data.

[0035] Based on the reflection information, a vegetation index is calculated, which is used to quantify the surface vegetation cover.

[0036] Based on the temporal changes in the vegetation index, regions where the vegetation index significantly decreased within the monitoring area were extracted through change vector analysis, forming a spectral feature layer.

[0037] Optionally, the step of synchronously acquiring surface polarization information collected by ground monitoring devices deployed within the monitoring area and processing it to generate a polarization feature map includes:

[0038] The Stokes vector is extracted from the surface polarization information, and the degree of linear polarization is calculated.

[0039] A threshold is set for the linear polarization degree, and pixels with a value higher than the threshold are identified as high-directional reflection regions, which correspond to the key regions with the reflective characteristics of the man-made object;

[0040] A polarization feature map is generated based on the highly directional reflection region.

[0041] Optionally, when the continuous change is determined to occur, generating the target violation patch and its corresponding geospatial information includes:

[0042] Potentially illegal patches that simultaneously meet the conditions of continuous decline in vegetation cover and significant expansion of man-made structures in the area are marked as target illegal patches.

[0043] Extract the set of contour boundary coordinates of the target violation patch in the geographic coordinate system of the monitoring area;

[0044] The set of contour boundary coordinates is associated with the unique identifier of the target illegal patch to generate geospatial information.

[0045] Optionally, the step of overlaying the associated digital orthophoto image and the boundary of the target illegal patch onto the monitoring video frame at the corresponding location based on the geospatial information of the target illegal patch includes:

[0046] Based on the geospatial information, the video stream from the surveillance camera is invoked, and the observation range of the surveillance camera covers the target violation patch;

[0047] The digital orthophoto image and the boundary of the target violation patch are superimposed onto the real-time image frame of the video stream; the digital orthophoto image is obtained by scheduling a drone to take pictures of the monitoring area.

[0048] Secondly, this application provides a monitoring system for a natural resource area, comprising:

[0049] The acquisition module is used to generate a spectral feature layer based on the acquired multi-source remote sensing data covering the monitoring area. The spectral feature layer is used to characterize the spectral feature differences of natural objects in the monitoring area as they transition from a first state to a second state.

[0050] It is also used to synchronously acquire surface polarization information collected by ground monitoring devices deployed in the monitoring area, and process it to generate a polarization feature map, which is used to distinguish the reflection characteristics of natural objects from the reflection characteristics of man-made objects;

[0051] The fusion module is used to fuse the spectral features of natural objects undergoing state changes identified from the spectral feature layer and the key regions with reflective properties of man-made objects identified from the polarization feature map, and to locate potential violation patches within the monitoring area;

[0052] The determination module is used to determine, based on the continuously acquired spectral feature layer and polarization feature map, whether the potential violation patch has undergone a continuous transition from a first state to a second state, wherein the determination of the transition needs to be based on both the transition of spectral features and the expansion of the key area with the man-made object.

[0053] The generation module is used to generate the target violation patch and its corresponding geospatial information when the continuous change is determined to occur.

[0054] The identification module is used to overlay the associated digital orthophoto image and the boundary of the target illegal patch onto the monitoring video screen at the corresponding location based on the geospatial information of the target illegal patch, and dynamically identify target activities within the boundary of the patch.

[0055] Thirdly, this application provides an electronic device, comprising:

[0056] Memory, used to store computer programs;

[0057] A processor is configured to execute the computer program to implement the steps of a method for monitoring a natural resource area as described in the first aspect above.

[0058] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of a natural resource area monitoring method as described in the first aspect above.

[0059] This application provides a method for monitoring natural resource areas. In this embodiment, by acquiring multi-source remote sensing data covering the monitoring area to generate a spectral feature layer, the spectral differences in the state changes of natural objects within the monitoring area can be clearly captured. By simultaneously acquiring the surface polarization information of the ground monitoring device and generating a polarization feature map, the reflectance characteristics of natural objects and man-made objects can be accurately distinguished, eliminating interference from man-made objects. By fusing key areas of the spectral features and polarization feature maps, potential illegal patches within the monitoring area can be accurately located. By determining whether potential illegal patches are continuously changing based on continuously acquired layers, misjudgment based on single data can be avoided, improving the accuracy of judgment. By generating target illegal patches and geospatial information after determining continuous changes, the specific location of the illegal area can be clearly identified. By overlaying digital orthophotos and patch boundaries onto the monitoring video and dynamically identifying target activities, real-time tracking of the situation in the illegal area can be achieved.

[0060] Furthermore, a change vector analysis algorithm is used to compare spectral feature layers from different time phases. Initial candidate areas with significantly reduced vegetation indices are identified using vegetation indices. Simultaneously, the linear polarization degree of surface polarization information is calculated, and pixels with high linear polarization degrees are selected to generate key areas. Then, a weighted decision fusion model is used to overlay and analyze the two types of areas. Overlapping areas that meet both conditions are clustered to locate potential violation patches. Through dual screening using vegetation indices and linear polarization degrees, combined with weighted fusion and clustering analysis, the selection range of potential violation patches is narrowed, eliminating misjudgments based on a single indicator and further improving the accuracy of potential violation patch location, ensuring that the selected areas better reflect actual violation scenarios. Attached Figure Description

[0061] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 A flowchart illustrating a method for monitoring a natural resource area, provided as an embodiment of this application;

[0063] Figure 2 This is a schematic diagram illustrating a specific implementation of a monitoring method for a natural resource area provided in this application.

[0064] Figure 3 This is a schematic diagram of the structure of a monitoring system for a natural resource area provided in an embodiment of this application. Detailed Implementation

[0065] In the monitoring of natural resource areas, the existing single optical remote sensing technology has obvious limitations. It cannot effectively distinguish between changes in the state of natural objects themselves and changes caused by human intervention, often resulting in misjudgments or omissions—either treating natural evolution as illegal behavior or ignoring illegal phenomena caused by the covert intervention of human objects. This leads to insufficient monitoring accuracy and makes it difficult to meet the needs of land management and ecological protection for precise monitoring.

[0066] To address the aforementioned issues, this application proposes a monitoring method for natural resource areas. The core of this method is the integration of multi-source remote sensing and ground monitoring data to achieve precise monitoring. Remote sensing data is used to generate a spectral feature layer to capture changes in the state of natural objects. This is combined with polarization information obtained from ground-based devices to distinguish the reflectivity of natural and man-made objects. Data fusion is then used to locate and continuously determine illegal patches, ultimately linking imagery and video for dynamic identification. This method overcomes the limitations of single-technology identification, solves the problems of false and missed detections at the source, and significantly improves the accuracy and reliability of natural resource violation monitoring.

[0067] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] The core of this application is to provide a method for monitoring natural resource areas, and a flowchart of one specific implementation method is shown below. Figure 1 As shown, the method includes:

[0069] S101. Generate a spectral feature layer based on the acquired multi-source remote sensing data covering the monitoring area.

[0070] The multi-source remote sensing data includes a first spectral image with a first spatial resolution and a second spectral image with a higher spatial resolution.

[0071] The spectral feature layer is used to characterize the differences in spectral features of natural objects within the monitoring area as they transition from a first state to a second state. These natural objects include vegetation and soil. The transition from the first state to the second state can be understood as a change from cultivated land to built-up land, or from dense vegetation to sparse vegetation, or from flat land to pits. Since these state transitions result in differences in the acquired spectral features, these differences can be used to construct the spectral feature layer.

[0072] Optionally, step S101 may specifically include the following steps:

[0073] S1011. Obtain the first spectral image of the monitored area through satellite remote sensing.

[0074] The first spectral image has a first spatial resolution and includes a combination of bands sensitive to vegetation and building materials. The band combination refers to the combination of bands within a specific wavelength range in the spectral image. Different bands have different sensitivities to objects such as vegetation and buildings. The red band is sensitive to the chlorophyll content of vegetation, while the near-infrared band is sensitive to the growth of vegetation.

[0075] In this embodiment of the application, a first spectral image covering the target monitoring area, such as a natural resource reserve, is captured by a spectral detection device carried on an in-orbit satellite. The image has a preset first spatial resolution and is filtered to include band combinations that are sensitive to vegetation and building materials.

[0076] In practical applications, a domestically produced satellite is selected to obtain coverage of 100km. 2 The first spectral image of the nature reserve was set with a first spatial resolution of 10 meters. The band combination selected was red light 650-680nm, near infrared 850-880nm and blue light 450-480nm. The red and near infrared bands are sensitive to vegetation, and the blue band helps to distinguish building materials.

[0077] S1012. Based on the first spectral image, identify suspicious areas where the change in land cover reaches a preset level.

[0078] Specifically, by combining the band combination information of the first spectral image with historical spectral images of the same period in the area, the spectral differences of the land cover are analyzed. When the difference value exceeds a preset threshold, the area is marked as a suspicious area, and subsequent monitoring is focused on this area.

[0079] For example, with a preset land cover spectral difference threshold of 0.2, comparing the first spectral images of the aforementioned natural resource protection area from 2025 and 2024 (the same period), the spectral difference value of each pixel is calculated, and pixels with a difference value ≥ 0.2 are grouped into 3. The areas were marked as suspicious, and it was initially determined that there may be changes in vegetation or land cover in these areas.

[0080] S1013. Dispatch the drone to take pictures of the suspicious area and obtain a second spectral image and a digital orthophoto map of the suspicious area. The second spectral image has a higher spatial resolution than the first spectral image.

[0081] In this step, the digital orthophoto map is obtained by geometrically correcting and stitching images taken by drones. It is an image map that can truly reflect the planar position and shape of surface features. The second spectral image has a higher resolution and can capture more subtle surface changes.

[0082] Specifically, based on the coordinate information of the suspicious area, the suspicious area is first identified using the first spectral image. A drone is then dispatched to the target area within the suspicious area to take pictures along a preset flight path, simultaneously acquiring a second spectral image and a digital orthophoto map. Utilizing the low-altitude shooting advantage of the drone, the spatial resolution of the image is improved, compensating for the lack of detail in the satellite image. A spectral feature layer is constructed based on the acquired second spectral image. Finally, the spectral feature layer and the polarization feature map are fused to locate potential violation patches within the suspicious area.

[0083] In practical applications, multi-rotor UAVs are deployed to cover the aforementioned 3km area. 2 Low-altitude imaging was conducted on the suspected area, with a pre-set flight path closely following the area's boundaries and key internal points. The spatial resolution of the second spectral image was set to 1 meter, significantly higher than the 10-meter resolution of satellite imagery. Simultaneously, a digital orthophoto map of the area was acquired. This not only clearly reveals the vegetation distribution and surface details within the suspected area but also provides reliable data support for subsequent construction of high-precision spectral feature layers and fusion of polarization feature maps to locate potential illegal patches.

[0084] S1014. Extract the reflection information of red light and near-infrared bands that are sensitive to vegetation status from the first and second spectral images of multi-source remote sensing data.

[0085] Among them, reflection information refers to the reflection intensity data of surface objects to light of different wavelengths. Vegetation reflects light weakly in the red band and strongly in the near-infrared band. This characteristic can be used to determine the status of vegetation.

[0086] In one specific implementation, firstly, the reflection intensity data corresponding to the red light 650-680nm and near-infrared 850-880nm bands are extracted from the first spectral image of the satellite and the second spectral image of the UAV, respectively. Then, the information directly related to the vegetation status is focused on, and irrelevant band interference is eliminated.

[0087] For example, from the satellite and drone images mentioned above, the red light band reflectance value and near-infrared band reflectance value of each pixel in the suspicious area can be extracted. The range of red light band reflectance value of satellite image pixels is 0.1-0.3, and the range of near-infrared band reflectance value is 0.4-0.7. Because of its higher resolution, drone imagery can capture the reflectance value differences of individual vegetation.

[0088] S1015. Based on the reflection information, the vegetation index is calculated, and the vegetation index is used to quantify the surface vegetation cover.

[0089] Specifically, the Normalized Difference Vegetation Index (NDVI) is used to calculate the vegetation index. This method quantifies vegetation cover and growth by comparing the ratio of red light to near-infrared reflectance. The NDVI ranges from -1 to 1, with negative values ​​typically corresponding to non-vegetated areas such as water bodies and buildings, 0 corresponding to bare soil, and positive values ​​corresponding to vegetated areas. A higher value indicates higher vegetation cover and better growth.

[0090] In practical applications, a pixel in a suspected area from a UAV image is selected. Its red band reflectance value is 0.2 and its near-infrared band reflectance value is 0.6. Substituting these values ​​into the normalized vegetation index (NDI) algorithm, the result is (0.6-0.2) / (0.6+0.2)=0.4 / 0.8=0.5, indicating that the area corresponding to this pixel has moderate vegetation cover. Another pixel has a red band reflectance value of 0.3 and a near-infrared band reflectance value of 0.3, with a calculation result of 0, indicating that this area is bare soil or has very little vegetation.

[0091] S1016. Based on the temporal changes in vegetation index, regions where vegetation index has significantly decreased within the monitoring area are extracted through change vector analysis, forming a spectral feature layer.

[0092] Among them, change vector analysis is a technique that quantifies the degree of surface change by calculating the magnitude and direction of changes in vegetation indices at different time series, while spectral feature layers present areas with significantly reduced vegetation indices and corresponding spectral differences in the form of images.

[0093] In this embodiment, the vegetation index of the same area is first compared with that of the same historical period. Then, the change in the index is calculated by change vector analysis. A change threshold is set, and the area where the vegetation index decreases by more than the threshold is extracted. Finally, the data are integrated to form a spectral feature layer, which intuitively marks the area where the vegetation status changes abnormally.

[0094] For example, a threshold of 0.3 was set for a significant reduction in vegetation index, compared to the above 3km 2The normalized vegetation index values ​​of the suspected area in 2025 and 2024 were compared. The change of each pixel was calculated by change vector analysis. Areas with a change of ≤-0.3 were extracted and integrated to form a spectral feature layer. This clearly marked the areas in the nature reserve where vegetation had been greatly reduced, providing precise targeting for locating potential illegal patches by fusing polarization feature maps.

[0095] This application utilizes a combination of technologies, including multi-source remote sensing data synergy from satellites and drones, band extraction and vegetation index calculation, and temporal change analysis, to accurately capture areas of abnormal vegetation changes. It leverages satellites to achieve large-scale monitoring coverage, while drones compensate for shortcomings in detail recognition, addressing the limitations of traditional single-source remote sensing data in terms of coverage or detail blurring, thereby improving the accuracy and efficiency of identifying surface vegetation changes.

[0096] S102. Simultaneously acquire the surface polarization information collected by the ground monitoring device deployed in the monitoring area, and process it to generate a polarization feature map.

[0097] Among them, surface polarization information refers to the polarization state data generated when ground objects reflect light. Different materials have different polarization characteristics of reflected light, which is the core basis for distinguishing between natural objects and man-made objects. The polarization feature map is an image formed by visualizing polarization information, which can intuitively present the differences in polarization reflection of different objects and is used to distinguish the reflection characteristics of natural objects from those of man-made objects.

[0098] Optionally, step S102 may specifically include the following steps:

[0099] S1021. Extract the Stokes vector from the surface polarization information and calculate the degree of linear polarization.

[0100] Among them, the Stokes vector is a physical quantity used to fully describe the polarization state of a light beam. It contains four components and can comprehensively characterize the polarization characteristics of light reflected from the Earth's surface. The degree of linear polarization is a parameter that measures the degree of linear polarization of a light beam and can quantitatively distinguish the polarization differences of light reflected from different objects.

[0101] In this embodiment, ground-based monitoring devices are used to collect surface polarization information within the monitoring area to ensure that the data covers the target monitoring range and is accurate and effective. Secondly, a polarization light analysis algorithm is used to extract the Stokes vector from the collected information to accurately capture the core features of the beam polarization state. Finally, the linear polarization degree is calculated based on this vector using the linear polarization degree formula, forming a complete data processing link.

[0102] In practical applications, taking a natural resource reserve as the monitoring scenario, the ground polarization monitoring equipment first collects the surface polarization information of a certain point in the area. Assuming that the Stokes vector is obtained through analytical algorithm processing (1.0, 0.3, 0.1, 0.0), the corresponding value is finally calculated by substituting it into the linear polarization degree formula. The calculation process involves first finding the sum of the squares of the two linear polarization components in the vector, i.e. The final linear polarization degree is approximately 0.316. The above example is merely one example of this application; in practical applications, settings can be made according to requirements, and this application does not limit this.

[0103] In another specific implementation, an optimized polarization analysis algorithm can be used to extract the Stokes vector, which is suitable for scenes with complex lighting conditions, can improve the accuracy of vector analysis, and thus optimize the calculation results of linear polarization degree.

[0104] S1022. Set a threshold for the linear polarization degree, and identify pixels with a value higher than the threshold as high-directional reflection areas. High-directional reflection areas correspond to key areas with the reflective properties of man-made objects.

[0105] Among these, man-made objects are not limited to building materials and other materials related to human perception; they can also be changes caused by excavation, such as the reflective properties of fresh soil, which differ from those of unexcavated land. High-directional reflectivity regions refer to areas where reflected light has strong directional polarization characteristics. Man-made objects, due to their uniform material and relatively smooth surfaces, reflect light with a higher degree of linear polarization, typically corresponding to high-directional reflectivity regions. The threshold, on the other hand, is a critical value determined based on extensive measured data, used to delineate high and low linear polarization regions.

[0106] In this embodiment, firstly, based on the previously measured linear polarization degree data of natural and man-made objects, the polarization degree distribution range of the two types of objects is statistically analyzed, and a reasonable threshold is determined; secondly, the linear polarization degree data of each pixel calculated in step S1021 is retrieved and compared with the set threshold one by one; finally, pixels with values ​​higher than the threshold are marked as high directional reflection areas, and these areas are initially determined to be suspicious areas where man-made objects exist, forming a complete process from data statistics to threshold comparison and then to area marking.

[0107] In practical applications, taking the same nature reserve as an example, firstly, based on the linear polarization degree data of natural objects such as vegetation and soil and man-made objects such as construction waste and temporary facilities in the area, a threshold of 0.25 is set; secondly, the linear polarization degree data of each pixel in the area is extracted, and it is assumed that the polarization degree of a certain pixel is found to be 0.316 through calculation; finally, it is compared with the threshold. Since the value is higher than 0.25, it is marked as a high directional reflection area, and it is determined that there may be man-made objects in the area.

[0108] S1023. Based on the high-directional reflection region, generate a polarization feature map.

[0109] Among them, image rendering technology is a technology that transforms data into visualized images. By giving different regions specific visual features, it makes data differences intuitively distinguishable. The core value of polarization feature maps lies in clearly distinguishing between areas where man-made objects may exist and areas where natural objects may exist.

[0110] Specifically, firstly, the linear polarization degree analysis results of all pixels in the monitoring area are integrated and associated with the geographic coordinate information of each pixel; secondly, a specific color is assigned to the marked high directional reflectance areas, and another differentiated color is assigned to the low linear polarization degree areas corresponding to natural objects to ensure that the two types of areas are clearly visually distinguishable; finally, the images after color marking are stitched and geometrically corrected to eliminate the distortion caused by the shooting angle and generate a complete and accurate polarization feature map.

[0111] For example, for the aforementioned nature reserve, firstly, color identification rules are determined, marking high-directional reflective areas with a linear polarization degree higher than 0.25 as red and natural object areas with a linear polarization degree lower than the threshold as green; secondly, color values ​​are assigned to all pixels to form a preliminary image; finally, after stitching and correction, a polarization feature map is generated, with red areas visually representing key areas where man-made objects may exist, and green areas corresponding to natural object areas.

[0112] This application utilizes a combination of Stokes vector analysis, linear polarization degree calculation, and threshold screening techniques to accurately distinguish the reflectance characteristics of natural and man-made objects. It solves the problem of traditional methods' difficulty in quickly identifying concealed man-made objects, overcomes the limitations of previous qualitative distinctions, and improves the accuracy and efficiency of distinguishing object reflectance characteristics. This represents a leap from qualitative to quantitative judgment in the field of natural resource monitoring, while achieving efficient identification without complex equipment, thus balancing practicality and accuracy.

[0113] S103. By integrating the spectral features of natural objects undergoing state changes identified from the spectral feature layer with key regions exhibiting reflective properties of man-made objects identified from the polarization feature map, potential violation patches are located within the monitoring area.

[0114] Among them, potential illegal patches refer to areas where there is a high probability of human-caused damage to natural resources. The core judgment basis is the superposition of the dual characteristics of abnormal changes in the state of natural objects and the intervention of man-made objects. Fusion processing is to conduct correlation analysis on feature data from two different dimensions to improve the accuracy of patch location.

[0115] Optionally, such as Figure 2 As shown, step S103 may specifically include the following steps:

[0116] S1031. Using a change vector analysis algorithm to compare spectral feature layers at different time phases, and by calculating the vegetation index representing the amount of vegetation cover change, pixels with significantly reduced vegetation index are identified as preliminary candidate regions for state change.

[0117] Among them, different time phases refer to data of the same area acquired at different time points. By comparing across time dimensions, dynamic changes in the land surface can be accurately captured; the vegetation index is the difference between the current vegetation index and the historical vegetation index for the same period, which can quantify the degree of change in vegetation cover; the preliminary candidate area is a suspicious area screened only based on spectral features, which needs to be further verified by combining polarization features.

[0118] In this embodiment, spectral feature layers of different time phases of the monitoring area are first retrieved, and the differences between the layers are compared using a change vector analysis algorithm. Then, the vegetation index of each pixel is calculated. Finally, a change threshold is set, and pixels with vegetation indices significantly lower than the threshold are selected and designated as preliminary candidate areas for state changes, laying the foundation for subsequent accurate positioning.

[0119] In practical applications, taking a natural resource reserve as a scenario, the spectral feature layers of the area in May 2025 and May 2024 were selected, the normalized vegetation index of each pixel was calculated, and the threshold for significant reduction was set to -0.3. Pixels with a change of ≤-0.3 were designated as preliminary candidate areas. These areas are likely to have changes in the state of natural objects such as vegetation destruction.

[0120] S1032. Simultaneously, the linear polarization degree is calculated based on the surface polarization information. Pixels with high linear polarization degree are selected by setting a threshold, and key areas representing the reflective properties of man-made objects are generated.

[0121] Among them, the key area is the area that concentrates the reflective properties of man-made objects. By filtering by linear polarization degree, the possible range of man-made objects can be accurately located. This is directly related to the high polarization degree of the light reflected by man-made objects, which can effectively distinguish between natural objects and man-made objects.

[0122] Specifically, firstly, the surface polarization information of the monitoring area is retrieved synchronously, and the linear polarization degree of each pixel is calculated using the method described above. Secondly, a linear polarization degree threshold is set in combination with the measured data. Finally, pixels with high linear polarization degree that are higher than the threshold are selected, integrated, and key areas are generated to clarify the possible distribution range of man-made objects.

[0123] For example, in the above-mentioned natural resource protection area scenario, the surface polarization information of the area is collected simultaneously, the linear polarization degree of each pixel is calculated, and combined with the measured data of the polarization characteristics of man-made objects and natural objects in the area, the high linear polarization degree threshold is set to 0.25. Pixels with polarization degree ≥ 0.25 are integrated into key areas. These areas are likely to contain man-made objects such as construction waste and temporary facilities.

[0124] S1033. Using a weighted decision fusion model, the preliminary candidate areas and key areas are spatially overlaid and analyzed. Overlapping areas that simultaneously meet the two conditions of significantly reduced vegetation index and high linear polarization are clustered to locate potential violation patches within the monitoring area.

[0125] In the above steps, the weighted decision fusion model is a multi-feature fusion analysis model that comprehensively determines regional attributes by assigning reasonable weights to different features, and the weights are calibrated based on the feature recognition accuracy. Spatial overlay analysis is a technique that associates the geographic coordinate information of two regions and filters the overlapping parts. Clustering processing aggregates scattered overlapping pixels into continuous patches, which facilitates subsequent positioning and monitoring.

[0126] In this embodiment, a weighted decision fusion model is first constructed. Model training requires the collection of a large number of measured samples, including spectral and polarization feature data of illegal and non-illegal regions. The weights of each feature are optimized with the goal of improving the recognition accuracy. The weights of spectral features and polarization features are set to 0.4 and 0.6, respectively. Secondly, the preliminary candidate regions and key regions are spatially superimposed to screen out overlapping pixels that meet both conditions. Finally, the scattered overlapping pixels are aggregated into continuous patches using a clustering algorithm, and the potential illegal patches are finally located.

[0127] In practical applications, a weighted decision fusion model is used to overlay analysis on the preliminary candidate areas and key areas of the aforementioned natural resource protection zones, to simultaneously meet the following criteria. Overlapping pixels with a polarization degree ≤ -0.3 and a linear polarization degree ≥ 0.25 are clustered into three consecutive patches using the K-means clustering algorithm. These patches are the potential illegal patches in the area, which are likely areas where facilities were built after vegetation was destroyed.

[0128] This application achieves precise superposition of natural object state changes and man-made object intervention features through a combination of technologies including change vector analysis, linear polarization degree screening, and weighted decision fusion. This solves the problem of misjudgment and omission in traditional single feature monitoring, and improves the accuracy and reliability of locating potential illegal patches.

[0129] S104. Based on the continuously acquired spectral feature layer and polarization feature map, determine whether the potential illegal patch has undergone a continuous transition from the first state to the second state. The determination of the transition needs to be based on both the change of spectral features and the expansion of the key area with man-made objects.

[0130] Among them, the continuous change in the state of the patch is not a random fluctuation, but a stable trend change in the time series; the dual judgment criteria can effectively eliminate the interference of natural fluctuations or temporary man-made objects, ensure the reliability of the judgment results, and avoid misjudging non-violation scenarios as violations.

[0131] Optionally, step S104 may specifically include the following steps:

[0132] S1041. For each potential violation patch, establish its spectral and polarization characteristic profile over a continuous time series; by performing trend analysis on the mean vegetation index within the patch, determine whether the vegetation cover of the potential violation patch shows a downward trend.

[0133] Among them, continuous time series refers to multiple periods of data collected at fixed time intervals, which can completely capture the dynamic evolution process of patch status; spectral and polarization feature archives are datasets that integrate multi-period feature data of a single patch, providing complete data support for trend analysis; the mean vegetation index is the average value of the vegetation index of all pixels in a single patch, which can weaken the impact of local abnormal pixels on the overall trend.

[0134] In this embodiment, firstly, for each potential illegal patch, spectral and polarization characteristic data of each period in the continuous time series are extracted to establish a unique feature file. Secondly, the mean vegetation index of each patch is calculated. Finally, a linear regression algorithm is used to perform trend analysis on the mean sequence, and the vegetation cover is determined to show a downward trend based on the regression results.

[0135] In practical applications, three potential violation patches in a natural resource reserve were used as the research subjects. Feature data from the past six months was collected monthly to create a database. The mean vegetation index for each patch was calculated for each period, assuming a sequence of [-0.32, -0.35, -0.34, -0.38, -0.40, -0.41]. Linear regression analysis of this sequence showed that when the regression coefficient was negative and the fit was good, the vegetation cover of the patch exhibited a continuous downward trend.

[0136] S1042. Calculate the area of ​​key regions with reflective properties of man-made objects in potential violation patches, and determine whether the area is expanding over time.

[0137] Among them, the key area of ​​man-made objects refers to the actual surface area covered by high linear polarization pixels within the map patch. By quantifying the distribution range of man-made objects through area, its time series changes can directly reflect whether man-made objects are continuously involved. The determination of area expansion needs to be combined with the set growth threshold to avoid judging small area fluctuations as trend expansion.

[0138] Specifically, firstly, based on the polarization feature map of the continuous time series, the key areas of man-made objects in each potential violation patch are extracted. Secondly, the actual area of ​​the key areas in each period is calculated by combining the spatial resolution of the image. Finally, the area data of adjacent periods and the overall sequence are compared to determine whether there is an expanding trend.

[0139] For example, in the aforementioned natural resource protection zone scenario, for the same potentially illegal patch, key areas of man-made structures are extracted based on monthly polarization feature maps. The area is calculated using a 1-meter spatial resolution, resulting in an area sequence of 20 square meters, 22 square meters, 25 square meters, 28 square meters, 31 square meters, and 35 square meters for the past six months. A significant area expansion threshold is set at an average monthly increase of ≥2 square meters. This sequence shows an average monthly increase of 2.5 square meters, indicating an expanding trend in the area of ​​key man-made structures. The above example is merely one illustration of this application; in practical applications, settings can be customized according to requirements, and this application does not impose any limitations on this.

[0140] S1043. A transformation process is confirmed as a continuous transformation if and only if a potential violation patch simultaneously meets both the criteria of a continuous decline in vegetation cover and a significant expansion in the area of ​​key man-made structures.

[0141] The core of the dual-condition judgment rule is to ensure the accuracy of the attribution of continuous changes. Only when both natural damage and man-made intervention are met can non-violation scenarios such as natural degradation and temporary stacking be excluded, and the continuous destructive behavior caused by humans be accurately identified.

[0142] In this embodiment of the application, the vegetation cover trend determination result of S1041 and the man-made area change determination result of S1042 are first summarized. Then, the two results are logically verified. Finally, only when both results are "yes" is it confirmed that the patch has undergone a continuous change. If either result is not satisfied, the possibility of a continuous change is excluded.

[0143] In practical applications, for potential illegal landfill patches in the aforementioned nature reserves, if the condition of continuous decline in vegetation cover is met, and the area of ​​key man-made structures increases by an average of 2.5 square meters per month, reaching a significant expansion threshold, and both conditions are satisfied, it is confirmed that the landfill has undergone a continuous transition from the first state to the second state. The above example is merely one example of this application; in practical applications, it can be set according to needs, and this application does not limit it in this regard.

[0144] This application achieves accurate determination of the changing trends of potential illegal patches by combining time series feature archiving, dual-dimensional trend judgment and logical verification. It solves the problems of traditional single-dimensional judgment being susceptible to interference and having a high misjudgment rate, ensuring that only scenarios of continuous human-caused damage are identified, and improving the rigor and reliability of violation judgment.

[0145] In addition, considering that remote sensing data acquisition takes a long time, and illegal construction activities usually take place at night, drones are used to collect nighttime light data of potential illegal areas.

[0146] Optionally, after locating a potential violation patch within the monitoring area, the step of determining whether the potential violation patch undergoes a continuous transition from a first state to a second state further includes:

[0147] Acquire nighttime light data covering the potential violation patch area; analyze the nighttime light data to identify whether there is an artificial light anomaly zone within the boundary of the potential violation patch that is created from nothing and whose brightness value and distribution range are stable or increase over time; use the existence of the artificial light anomaly zone as an additional verification condition for determining the occurrence of the continuous transition; when a potential violation patch simultaneously satisfies the three conditions of continuous transition of spectral characteristics, expansion of key areas with man-made objects, and existence of the artificial light anomaly zone, it is determined that a continuous transition from the first state to the second state has occurred.

[0148] In the above steps, nighttime light data refers to the data such as light radiation intensity and distribution range collected at night in potential illegal patch areas using night vision imaging equipment mounted on drones; artificial light anomaly areas refer to artificial light source areas that appear within the patch boundaries and are different from natural light sources. The core judgment characteristic is that the light appears from nothing and the brightness value and distribution range show a stable or increasing trend over time; additional verification conditions refer to supplementary verification criteria added on the basis of the original dual judgment conditions, which are used to enhance the accuracy of continuous change judgment and make up for the shortcomings of long remote sensing data cycles and weak nighttime monitoring.

[0149] Specifically, after locating potential illegal patches in the monitoring area, drones equipped with night vision imaging devices are deployed to collect light data on the potential illegal patch areas at night, based on the geospatial information of the patches, ensuring that the collection range completely covers the boundaries of the patches and the surrounding areas. Secondly, the nighttime light data collected over several consecutive nights is preprocessed to remove interference from natural light sources such as moonlight and starlight, extracting valid artificial light signals. The brightness, distribution range, and variation patterns of the lights at different times are analyzed to identify any abnormal artificial light areas that have appeared from scratch and whose indicators are stable or increasing. Finally, the existence of abnormal artificial light areas is used as an additional verification condition, combined with two other conditions: a continuous decline in the original vegetation cover and an expansion of key areas of man-made structures. Only when all three conditions are met simultaneously can it be determined that a potential illegal patch has undergone a continuous transition from the first state to the second state.

[0150] In another specific implementation, a multi-time-period layered acquisition strategy can be adopted to collect light data at different times such as early morning and late at night, avoiding the randomness of a single acquisition. At the same time, the light spectral characteristics can be combined to further distinguish different types of light sources such as construction lighting and temporary lighting, thereby improving the accuracy of identifying abnormal artificial lighting areas.

[0151] In practical application, taking the potentially illegal area TG-001 in an ecological reserve as an example, drones were deployed to collect nighttime light data of the area for seven consecutive nights from 22:00 to 4:00 the following morning. The data collection frequency was once per night, and the onboard night vision equipment had a resolution of 1 meter, capable of capturing light signals with a brightness range of 0-255. After preprocessing to remove interference from natural light sources, data analysis revealed that: there were no artificial light signals within the boundary of the area in the first two nights; sporadic lights appeared on the third night, with a brightness value of 80 and a distribution area of ​​approximately 5 square meters; in the following four nights, the brightness value gradually increased to 150, and the distribution area expanded to 20 square meters, with both brightness and range maintaining a stable upward trend, successfully identifying an abnormal area of ​​artificial lighting. Combining the original conditions, the area already met the additional conditions of continuously declining vegetation cover, expansion of key man-made areas, and the presence of an abnormal area of ​​artificial lighting, ultimately determining that it was undergoing a continuous transformation.

[0152] This application addresses the issues of long acquisition cycles for remote sensing data and blind spots in the monitoring of illegal nighttime activities by adding a nighttime light data collection and analysis step. It effectively eliminates interference factors such as temporary activities and natural fluctuations, and significantly improves the rigor and accuracy of continuous change determination.

[0153] S105. When a continuous change is determined, generate the target violation patch and its corresponding geospatial information.

[0154] Among them, the target violation patch refers to the patch that has been verified by dual conditions to have continuous human-caused damage, which is different from the previous potential violation patch and has a clear violation orientation; geospatial information is a data set of geographic attributes such as the location and boundaries of the associated patch.

[0155] Optionally, step S105 may specifically include the following steps:

[0156] S1051. Potentially illegal patches that simultaneously meet the two conditions of continuous decline in vegetation cover and significant expansion of man-made structures in the area are marked as target illegal patches.

[0157] The core of the marking operation is to define the target from "potentially suspicious" to "clear target". Through double-condition secondary verification, potential patches that are misjudged or missed are eliminated to ensure the accuracy of the target violation patches. The marking results need to be linked with the previous feature files to retain complete judgment basis.

[0158] In this embodiment of the application, the results of the dual-condition judgment in S104 are first summarized, and potential illegal patches that meet both conditions are selected. Then, a unique identifier is assigned to these patches using a special marking rule to distinguish them from potential patches that do not meet the conditions. Finally, the marking results are bound to the patch feature file to form complete basic data of the target patches.

[0159] In practical application, taking three potential violation patches in a nature reserve as examples, after verification, only one patch simultaneously met the conditions of continuous decline in vegetation cover and significant expansion of the key area of ​​man-made structures. According to the rules, it was assigned the marking code "TG-001" and linked with the characteristic data and judgment records of the previous six months, officially marking it as the target violation patch. The above example is only one example of this application; in practical applications, it can be set according to needs, and this application does not limit this.

[0160] S1052. Extract the set of contour boundary coordinates of the target illegal patch in the geographic coordinate system of the monitoring area.

[0161] Among them, the geographic coordinate system is a reference system used to locate the position on the ground surface, which can convert the image pixel position of the patch into the real geographic coordinates; the outline boundary coordinate set refers to a series of continuous coordinate points that constitute the boundary of the patch. The spatial range of the patch is accurately delineated through the coordinate set to ensure accurate positioning.

[0162] Specifically, firstly, the spectral feature layer and polarization feature map corresponding to the target violation patch are retrieved to determine the pixel boundary of the patch in the image. Secondly, combined with the geographic coordinate system parameters used in the monitoring area, the pixel boundary is converted into geographic coordinates through a coordinate transformation algorithm. Finally, continuous coordinate points on the boundary are extracted to form a set of contour boundary coordinates.

[0163] For example, in the above-mentioned natural resource protection area scenario, the target illegal patch “TG-001” uses the WGS-84 geographic coordinate system. By using a coordinate transformation algorithm, its outline boundary coordinate set is extracted, resulting in a series of continuous coordinate points (118.XXXXX, 32.XXXXX), (118.XXXXX, 32.XXXXX)... a total of 20 coordinate points, which completely delineate the real spatial boundary of the patch.

[0164] S1053. Associate the set of contour boundary coordinates with the unique identifier of the target illegal patch to generate geospatial information.

[0165] Among them, the unique identifier is a special code used to distinguish different target illegal patches, ensuring that the geospatial information of each patch can be retrieved and associated separately; the generated geospatial information is comprehensive data that integrates patch identifiers, boundary coordinates, spatial range and other attributes, realizing the unification of patch "identity information" and "location information".

[0166] In this embodiment, the unique identifier and contour boundary coordinate set of the target illegal patch are first retrieved. Then, the two are bound together by a data association algorithm to supplement the patch area, the location of the patch, and other derived spatial attributes. Finally, the associated data is formatted to generate a standardized geospatial information dataset.

[0167] In practical applications, for the target violation patch "TG-001", its unique identifier is associated with a set of 20 extracted boundary coordinate points. Additional attributes, such as the calculated patch area of ​​35 square meters and its sub-region, are then used to generate formatted geospatial information, which can be directly used for map loading and on-site navigation and positioning. The above example is merely one illustration of this application; in practical applications, it can be customized according to requirements, and this application does not impose any limitations on this.

[0168] This application achieves the accurate generation of target illegal patches and geospatial information through a combination of dual-condition screening and marking, precise coordinate extraction and data association. It solves the problems of vague positioning and disconnect between identity and location of traditional illegal patches, ensuring that illegal patches can be accurately traced and controlled.

[0169] S106. Based on the geospatial information of the target illegal patch, the associated digital orthophoto image and the boundary of the patch corresponding to the target illegal patch are overlaid into the monitoring video image at the corresponding location, and the target activity is dynamically identified within the patch boundary.

[0170] Among them, the overlay operation refers to the precise alignment of static images and boundary data with dynamic video footage to achieve spatial dimension superposition and fusion; the core of dynamic target activity recognition is to capture the moving targets and behaviors of people, equipment and other objects within the map patch, and through the precise identification and behavior analysis of construction machinery, to form a preliminary judgment on whether there is a violation and trigger the corresponding verification process.

[0171] Optionally, step S106 may specifically include the following steps:

[0172] S1061. Based on geospatial information, call the video stream from the surveillance camera, and the observation range of the surveillance camera covers the target illegal patch.

[0173] Among them, video stream refers to the continuous image data transmitted in real time by the surveillance camera, which can realize real-time visual monitoring of the target area; the core of camera matching is to filter the equipment that covers the target illegal patches based on geospatial information to ensure that the activities within the patches can be completely captured.

[0174] In this embodiment, the geospatial information of the target illegal patch is first parsed to determine its spatial range and coordinate interval. Then, the location and observation range parameters of all surveillance cameras in the monitoring area are retrieved, and the cameras covering the coordinate interval are selected. Finally, the real-time video stream of the corresponding camera is called through the device interface.

[0175] In practical applications, for the illegal landfill TG-001 in the natural resource protection area, whose geographical coordinate range is (118.XXXXX-118.XXXXX, 32.XXXXX-32.XXXXX), there are suspected signs of illegal construction at night in this area. The high-definition night vision camera with the number CAM-028 was found to have a completely covered range and has the ability to clearly identify construction machinery. The 1080P real-time video stream of the camera was successfully called through the system interface, with the frame rate maintained at 25 frames per second, ensuring that the movement details of equipment such as excavators can be accurately captured.

[0176] S1062. The digital orthophoto image and the boundary of the target violation patch are superimposed onto the real-time image frame of the video stream; the digital orthophoto image is obtained by scheduling a drone to take pictures of the monitoring area.

[0177] Among them, real-time image frames refer to continuous single-frame images in the video stream, which are the basic carriers for overlay operations. Overlay processing must ensure that the spatial accuracy of the image, boundary and video screen is consistent, and ensure that the boundary of the image patch can accurately define the target area in the video without offset or misalignment.

[0178] Specifically, firstly, the digital orthophoto image and the set of contour boundary coordinates corresponding to the target violation patch are extracted. Secondly, the coordinates of the real-time image frames of the video stream are calibrated to keep them consistent with the geographic coordinate system. Finally, an image overlay algorithm is used to overlay the digital orthophoto image as a base map and the patch boundary in the form of a conspicuous line onto each frame of the image, thereby achieving the fusion of static data and dynamic images.

[0179] For example, in the aforementioned natural resource protection area scenario, the digital orthophoto of the UAV corresponding to TG-001 and the red highlighted outline boundary are superimposed onto the real-time image frame of the CAM-028 camera. Coordinate calibration ensures that the boundary accurately defines the suspected construction area. After superposition, the image details are not obscured, and it can clearly define whether equipment such as excavators are active within the illegal plot, making it easy to accurately track their operating range.

[0180] S1063. Extract the pixel region located within the boundary of the patch from continuous surveillance video images.

[0181] In this context, a pixel region refers to the set of all pixels within a video frame bounded by a patch boundary. The core of the extraction operation is to focus on the target area, eliminate interference from irrelevant images outside the boundary, reduce subsequent data processing, and improve recognition efficiency. The surveillance video footage is acquired through tower monitoring cameras.

[0182] In this embodiment, the pixel coordinate range corresponding to the boundary of the patch in each frame image is first determined. Then, a region extraction algorithm is used to filter out all pixels within the range to form an independent pixel region dataset. Finally, the extracted pixel regions are preprocessed to remove noise pixels and ensure region integrity.

[0183] In practical applications, for the real-time image frames of the CAM-028 camera with a resolution of 1920×1080, the pixel coordinate range corresponding to TG-001 is (500-800, 300-600). A total of 90,000 pixels within this range are extracted to form a pixel region. After noise removal, 89,800 effective pixels are retained. This region is focused on to accurately capture the core features of the excavator, such as its outline and texture, and to avoid irrelevant objects outside the boundary from interfering with the recognition results.

[0184] S1064. Compare the pixel value differences in pixel regions of adjacent video frames to obtain the difference regions representing changes in pixel values.

[0185] In this step, pixel value difference refers to the change in brightness, color and other values ​​of pixels at the same position in two adjacent frames. The difference area is the set of pixels whose pixel value changes exceed a set threshold, which can intuitively reflect whether there is a moving target in the picture and the target position.

[0186] In this embodiment, pixel regions extracted from two adjacent frames are first selected, then the difference between pixels at corresponding positions is calculated, a pixel value difference threshold is set, and finally pixels with differences exceeding the threshold are selected, integrated to form a difference region, and the changed parts in the image are marked.

[0187] For example, pixel regions of two adjacent frames of images from the CAM-028 camera are selected, the gray value difference of each corresponding pixel is calculated, the difference threshold is set to 20, the gray value range is 0-255, and pixels with gray value differences ≥20 are selected and integrated to form an irregular difference region. The outline of this region roughly matches the body of the excavator, and it is preliminarily determined that there is a moving target such as an excavator in the image patch.

[0188] S1065. Based on the relationship between the position and shape of the difference region in the picture in multiple consecutive frames, determine at least one independently moving target and track the continuous position of the moving target in the picture.

[0189] Among them, the correlation relationship refers to the correlation features such as position offset and shape similarity of the difference regions in consecutive frames, which can distinguish independent moving targets from interfering changes; moving target tracking is to lock the target through the algorithm and continuously record its position change trajectory to ensure that the target is not lost.

[0190] In one specific implementation, firstly, the difference regions in 10 consecutive frames of images are extracted, and the position offset and shape similarity of the difference regions in each frame are analyzed. The difference regions with high correlation are classified as the same moving target, and independent moving targets are distinguished. Secondly, the Kalman filter algorithm is used to predict the target's position in the next frame, and the position is corrected in combination with the actual detection results to achieve continuous tracking of the moving target.

[0191] In practical applications, analyzing video footage from the aforementioned natural resource protection area across 10 consecutive frames revealed a set of correlated difference regions with stable outlines and slow-moving positions. The shape of these regions closely matched the excavator's body, identifying the excavator as a single, independent moving target. A Kalman filter algorithm was used to track its position, recording continuous coordinates (550, 350), (553, 355), (556, 360), etc., to accurately capture the excavator's slow-moving trajectory within the map area, achieving stable tracking. The above example is merely one illustration of this application; further modifications are possible based on specific needs in practical applications, and this application does not impose any limitations on this.

[0192] S1066. Extract the shape, texture and motion pattern features of the moving target in the video frame, and classify the moving target into a predefined dynamic target type based on the preset feature comparison rules.

[0193] Among them, dynamic target types refer to predefined target categories such as personnel, machinery, and vehicles, and their feature library contains typical shapes, textures, and motion patterns of various targets; feature comparison rules are based on the judgment criteria of feature similarity to ensure accurate classification.

[0194] In this embodiment, the shape features of each moving target are first extracted: contour size and aspect ratio; texture features: surface grayscale distribution; and motion pattern features: movement speed and trajectory pattern. Then, a preset dynamic target feature library is retrieved, and a feature similarity comparison algorithm is used to calculate the similarity between the target features and the features of each category in the library. Finally, the moving target is classified into the corresponding dynamic target type based on the similarity results.

[0195] In practical applications, the features of the aforementioned moving target are extracted. Its aspect ratio is approximately 4:1, it has a distinct bucket protrusion, and its texture exhibits a rough, granular feel reminiscent of tracks. Its movement pattern is low-speed, uniform movement with intermittent pauses, corresponding to a shoveling action. The similarity to the "excavator" category in the feature library reaches 92%, far exceeding other target categories, and it is ultimately accurately classified as an excavator. The feature similarity calculation can employ the cosine similarity formula, which will not be elaborated upon in this application.

[0196] S1067. Analyze the change pattern of the continuous position of the tracked moving target over time, and determine the behavior category of the moving target based on the preset behavior pattern rules.

[0197] Among them, the location change pattern refers to the target's movement trajectory, dwell time, activity range and other characteristics, which are the core basis for behavior judgment; the behavior category refers to predefined behaviors such as dwelling and observation, operation and construction, and round-trip transportation, which are adapted to the judgment needs of different violation scenarios.

[0198] Specifically, firstly, the continuous position data of the moving target is summarized, its trajectory, dwell time and activity range within the patch are analyzed, and the position change pattern is extracted. Secondly, the preset behavior pattern rule base is retrieved, and the extracted pattern is compared with various behavior features in the rule base to determine the behavior category of the moving target.

[0199] For example, in the above-mentioned natural resource protection area scenario, analyzing the excavator's position change pattern, it was found that it moved slowly in the densely vegetated area within the map patch, stayed for more than 40 minutes, and its trajectory showed a reciprocating digging pattern, accompanied by intermittent bucket lifting and lowering movements. This fully matches the behavioral characteristics of "excavator illegally digging vegetation" in the rule base, and it is determined that the excavator is engaging in illegal operation.

[0200] S1068. When at least one of the dynamic target type and behavior category does not meet the preset violation judgment conditions, a task awaiting manual verification is generated.

[0201] In this step, the violation judgment conditions refer to the predefined allowed target types and behavior ranges. Exceeding these ranges triggers a verification task. The task awaiting manual verification is a task list that integrates target information, behavior information, and video clips, providing a complete basis for manual review and avoiding machine misjudgment.

[0202] In this embodiment, the system first retrieves the preset violation judgment conditions and verifies the dynamic target type and behavior category of the moving target. If any item does not meet the conditions, the system automatically collects target tracking segments, feature data, and location information, generates standardized tasks to be manually checked, assigns them to the corresponding checkers, and finally records the task generation time and related information to facilitate tracking the check progress.

[0203] In practical application, the violation determination criteria for the aforementioned natural resource protection area are prohibiting excavators from entering the core protection area and prohibiting the excavation that damages vegetation. Upon verification, the dynamic target type is an excavator, and the behavior category is illegal excavation of vegetation. Neither of these criteria meets the preset conditions. The system automatically collects a one-minute high-definition video clip of the excavator's illegal operation, a feature recognition report, and precise location information, generating a task numbered HC-001 for manual verification. This task is immediately assigned to on-site verification personnel for rapid verification and handling. The above example is merely one example of this application. In practical applications, settings can be customized according to requirements, and this application does not limit this.

[0204] In another specific implementation, a tiered triggering rule can be adopted, which divides tasks into two levels, general and urgent, based on the severity of the violation. Urgent tasks are given priority in allocation, thereby improving the timeliness of verification and handling.

[0205] This application breaks down the barriers between static patch location and dynamic target monitoring by combining technologies such as video stream calling, image overlay, dynamic recognition and behavior determination. It overcomes the limitations of single machine recognition and achieves accurate locking and verification of illegal targets and behaviors, realizing progress from static control to dynamic closed-loop control in the field of natural resource monitoring.

[0206] Figure 3 This is a schematic diagram illustrating a specific implementation of a monitoring system for natural resource areas provided in this application. (Refer to...) Figure 3 The system may include:

[0207] The acquisition module 31 is used to generate a spectral feature layer based on the acquired multi-source remote sensing data covering the monitoring area. The spectral feature layer is used to characterize the spectral feature differences of natural objects in the monitoring area as they transition from a first state to a second state.

[0208] It is also used to synchronously acquire surface polarization information collected by ground monitoring devices deployed in the monitoring area, and process it to generate polarization feature maps. The polarization feature maps are used to distinguish the reflection characteristics of natural objects from those of man-made objects.

[0209] The fusion module 32 is used to fuse the spectral features of natural objects undergoing state changes identified from the spectral feature layer and the key regions with reflective properties of man-made objects identified from the polarization feature map, so as to locate potential violation patches within the monitoring area.

[0210] The determination module 33 is used to determine whether a potential illegal patch has undergone a continuous transition from a first state to a second state based on the continuously acquired spectral feature layer and polarization feature map. The determination of the transition needs to be based on both the change in spectral features and the expansion of the key area containing the man-made object.

[0211] The generation module 34 is used to generate the target violation patch and its corresponding geospatial information when the determination shows a continuous change.

[0212] The identification module 35 is used to overlay the associated digital orthophoto image and the boundary of the target illegal patch onto the monitoring video screen at the corresponding location based on the geospatial information of the target illegal patch, and dynamically identify target activities within the patch boundary.

[0213] The monitoring system for natural resource areas in this application is used to implement the aforementioned monitoring method for natural resource areas. Therefore, the specific implementation of the monitoring system for natural resource areas can be found in the embodiment section of the monitoring method for natural resource areas above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0214] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the natural resource area monitoring method described above.

[0215] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described methods for monitoring natural resource areas.

[0216] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0217] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the natural resource area monitoring method.

[0218] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0219] The above provides a detailed description of a monitoring method and system for natural resource areas provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for monitoring natural resource areas, characterized in that, include: Based on the acquired multi-source remote sensing data covering the monitoring area, a spectral feature layer is generated. The spectral feature layer is used to characterize the spectral feature differences of natural objects in the monitoring area as they transition from a first state to a second state. Simultaneously acquire surface polarization information collected by ground monitoring devices deployed within the monitoring area, and process it to generate a polarization feature map, which is used to distinguish the reflection characteristics of natural objects from those of man-made objects; By fusing the spectral features of natural objects undergoing state changes identified from the spectral feature layer with key regions exhibiting reflective properties of man-made objects identified from the polarization feature map, potential violation patches are located within the monitoring area; Based on the continuously acquired spectral feature layer and polarization feature map, it is determined whether the potential violation patch has undergone a continuous transition from the first state to the second state, wherein the determination of the transition needs to be based on both the change in spectral features and the expansion of the key area with the man-made object. When the continuous change is determined to occur, a target violation patch and its corresponding geospatial information are generated. Based on the geospatial information of the target illegal patch, the associated digital orthophoto image and the patch boundary corresponding to the target illegal patch are overlaid into the monitoring video image at the corresponding location, and the target activity is dynamically identified within the patch boundary; The step of determining whether a potential violation patch has undergone a continuous transition from a first state to a second state after locating it within the monitoring area further includes: Acquire nighttime light data covering the areas of the potential violation patches; Analyze the nighttime light data to identify whether there are any artificial light anomaly areas within the boundaries of the potential violation patches that appear from nothing and whose brightness values ​​and distribution ranges remain stable or increase over time. The presence of the aforementioned abnormal artificial lighting area is used as an additional verification condition for determining that the continuous change has occurred. When a potential violation patch simultaneously meets three conditions—a continuous shift in spectral characteristics, an expansion of a key area containing artificial objects, and the presence of an abnormal area of ​​artificial lighting—a continuous shift from the first state to the second state is determined to have occurred.

2. The method according to claim 1, characterized in that, The fusion of spectral features of natural objects undergoing state changes identified from the spectral feature layer and key regions with reflective properties of man-made objects identified from the polarization feature map, to locate potential violation patches within the monitoring area, includes: By comparing spectral feature layers from different time periods using a change vector analysis algorithm, and calculating the vegetation index that represents the amount of vegetation cover change, pixels with significantly reduced vegetation indexes are identified as preliminary candidate regions for state change. Simultaneously, the linear polarization degree is calculated based on the surface polarization information, and pixels with high linear polarization degree are selected by setting a threshold to generate key areas representing the reflective properties of man-made objects; By using a weighted decision fusion model, the preliminary candidate region and the key region are spatially overlaid and analyzed. Overlapping regions that simultaneously meet the two conditions of significantly reduced vegetation index and high linear polarization are clustered to locate potential violation patches within the monitoring area.

3. The method according to claim 2, characterized in that, The determination of whether the potential violation patch undergoes a continuous transition from a first state to a second state based on the continuously acquired spectral feature layer and polarization feature map includes: For each potential violation patch, a spectral and polarization characteristic profile is established over a continuous time series; by performing trend analysis on the mean vegetation index within the patch, it is determined whether the vegetation cover of the potential violation patch shows a downward trend. Calculate the area of ​​key regions with reflective properties of man-made objects in the potential violation patches, and determine whether the area is increasing over time. A transformation process is considered a continuous transformation if and only if a potential violation patch simultaneously meets both criteria: a continuous decline in vegetation cover and a significant expansion in the area of ​​key man-made structures.

4. The method according to claim 1, characterized in that, The dynamic identification of target activities within the boundary of the patch includes: Extract the pixel region located within the boundary of the patch from the continuous surveillance video footage; By comparing the pixel value differences in the pixel region in adjacent video frames, the difference region representing the pixel value change is obtained; Based on the correlation between the position and shape of the difference region in the image in multiple consecutive frames, at least one independently moving target is determined, and the continuous position of the moving target in the image is tracked. Extract the shape, texture and motion pattern features of the moving target in the video frame, and classify the moving target into a predefined dynamic target type based on preset feature comparison rules; Analyze the change pattern of the continuous position of the tracked moving target over time, and determine the behavior category of the moving target according to the preset behavior pattern rules; When the dynamic target type and at least one of the behavior categories do not meet the preset violation judgment conditions, a task awaiting manual verification is generated.

5. The method according to claim 1, characterized in that, The generated spectral feature layer includes: Reflectance information in the red and near-infrared bands, which are sensitive to vegetation status, is extracted from the first and second spectral images of the multi-source remote sensing data. Based on the reflection information, a vegetation index is calculated, which is used to quantify the surface vegetation cover. Based on the temporal changes in the vegetation index, regions where the vegetation index significantly decreased within the monitoring area were extracted through change vector analysis, forming a spectral feature layer.

6. The method according to claim 1, characterized in that, The process of synchronously acquiring surface polarization information collected by ground monitoring devices deployed within the monitoring area and processing it to generate a polarization feature map includes: The Stokes vector is extracted from the surface polarization information, and the degree of linear polarization is calculated. A threshold is set for the linear polarization degree, and pixels with a value higher than the threshold are identified as high-directional reflection regions, which correspond to the key regions with the reflective characteristics of the man-made object; A polarization feature map is generated based on the highly directional reflection region.

7. The method according to claim 1, characterized in that, When the continuous change is determined to occur, the generation of the target violation patch and its corresponding geospatial information includes: Potentially illegal patches that simultaneously meet the conditions of continuous decline in vegetation cover and significant expansion of man-made structures in the area are marked as target illegal patches. Extract the set of contour boundary coordinates of the target violation patch in the geographic coordinate system of the monitoring area; The set of contour boundary coordinates is associated with the unique identifier of the target illegal patch to generate geospatial information.

8. The method according to claim 1, characterized in that, The step of overlaying the associated digital orthophoto image with the boundary of the target illegal patch onto the monitoring video frame at the corresponding location based on the geospatial information of the target illegal patch includes: Based on the geospatial information, the video stream from the surveillance camera is invoked, and the observation range of the surveillance camera covers the target violation patch; The digital orthophoto image and the boundary of the target violation patch are superimposed onto the real-time image frame of the video stream; the digital orthophoto image is obtained by scheduling a drone to take pictures of the monitoring area.

9. A monitoring system for a natural resource area, characterized in that, include: The acquisition module is used to generate a spectral feature layer based on the acquired multi-source remote sensing data covering the monitoring area. The spectral feature layer is used to characterize the spectral feature differences of natural objects in the monitoring area as they transition from a first state to a second state. It is also used to synchronously acquire surface polarization information collected by ground monitoring devices deployed in the monitoring area, and process it to generate a polarization feature map, which is used to distinguish the reflection characteristics of natural objects from the reflection characteristics of man-made objects; The fusion module is used to fuse the spectral features of natural objects undergoing state changes identified from the spectral feature layer and the key regions with reflective properties of man-made objects identified from the polarization feature map, and to locate potential violation patches within the monitoring area; The determination module is used to determine, based on the continuously acquired spectral feature layer and polarization feature map, whether the potential violation patch has undergone a continuous transition from a first state to a second state, wherein the determination of the transition needs to be based on both the transition of spectral features and the expansion of the key area with the man-made object. The generation module is used to generate the target violation patch and its corresponding geospatial information when the continuous change is determined to occur. The identification module is used to overlay the associated digital orthophoto image and the boundary of the target illegal patch onto the monitoring video screen at the corresponding location based on the geospatial information of the target illegal patch, and dynamically identify target activities within the boundary of the patch.

Images